AUDITOR COMPENSATION AND AUDIT FAILURE: AN EMPIRICAL ANALYSIS* Mukesh Bajaj Managing Director LECG, LLC.

Transcription

1 AUDITOR COMPENSATION AND AUDIT FAILURE: AN EMPIRICAL ANALYSIS* by Mukesh Bajaj Managing Director LECG, LLC. Katherine Gunny University of California - Berkeley and Atulya Sarin Professor of Finance Santa Clara University February 27, 2003 * We are grateful for helpful comments received from David Denis. Sarin acknowledges support from Dean Witter Foundation. Please address all correspondence to Atulya Sarin (Ph: Asarin@scu.edu).

2 AUDITOR COMPENSATION AND AUDIT FAILURE: AN EMPIRICAL ANALYSIS Abstract Record number of audit failures during the recent past has prompted much debate about whether high auditor compensation, especially for nonaudit work, may have led to lax auditing standards. We shed light on this question by comparing auditor compensation for a set of firms in which accounting improprieties were alleged in a shareholder class action lawsuit with a set of matched firms in the same industry and of similar size. Our evidence suggests that auditors were not compensated differently for either their audit or consulting services over the period in which their client was allegedly involved in an accounting fraud. However, for the set of firms with the largest market reaction to the alleged fraud, the nonaudit component of the total fees was significantly higher than comparable firms even after controlling for other known determinants of auditor compensation.

3 AUDITOR COMPENSATION AND AUDIT FAILURE: AN EMPIRICAL ANALYSIS 1. Introduction The recent collapse of the stock market has been accompanied by large-scale audit failures. 1 Companies like Xerox, Enron and WorldCom, amongst others, have disclosed improprieties in their financial statements in the amounts of billions of dollars. Xerox disclosed that it had incorrectly realized $6.4 billion in revenues and overstated its pretax income by $1.41 billion over five years from 1997 to In some of these high profile audit failures, auditor misconduct has been alleged. For example, according to the Wall Street Journal dated January 23, 2003, the SEC is set to file civil fraud charges against KPMG for its role in auditing Xerox. The accounting improprieties at Enron regarding related party transactions led not only to its demise but also the failure of its auditor, Arthur Andersen. Record number of audit failures during the recent past, especially at some of the most well known and valuable firms in Corporate America has prompted much debate in the popular press, by the SEC, and in the US Congress about whether high auditor compensation, especially for nonaudit work, may have led to lax auditing standards. Three recent legislative reforms have attempted to address the perceived lax auditing due to auditors being compromised in their bid to attract consulting business from their auditing clients. First, the Sarbanes-Oxley Act of 2002, in an effort to improve audit quality, prohibits auditing firms and their personnel from providing any nonaudit services to auditing clients contemporaneously with the audit unless the additional services are preapproved by the company s audit committee. Second, in June 2002 the SEC banned auditors from performing nonaudit services in nine specific areas that may impair independence. Also, in an 1 In 2002 a record number of restatements erased billions of dollars of previously recorded revenue from financial statements. In 2002 restatements rose 22% from 2001 and restatements of large corporations (with annual revenue over 1 billion), 74 total, was almost double that of The Wall Street Journal, Restatements Rise 22%, 1/21/

4 effort to help investors assess the independence of a firm s auditor the SEC has changed the disclosure requirements regarding auditor compensation. On February 6, 2003 the SEC issued Final Rule: Strengthening the Commission's Requirements Regarding Auditor Independence which adopts amendments to its existing requirements in an effort to enhance the independence of the accountants. The amendment has three major changes to the existing rules. First, it increases the disclosure categories of professional fees paid for audit and nonaudit services from three to four: audit fees, audit-related fees, tax fees, and all other fees. Second, the new disclosure would require firms to report fees for each of the two most recent fiscal years. Third, the definition of audit fee has expanded to not only include services necessary to perform an audit in accordance with Generally Accepted Auditing Standards (GAAS) but also may include services that generally only the independent accountant reasonably can provide, such as comfort letters, statutory audits, attest services, consents and assistance with and review of documents filed with the Commission. 23 Although the SEC has enacted legislation assuming consulting assignments impair auditor independence, the evidence is mixed. On one hand it is argued that there are economies of scale and scope that permit the auditors to perform important consulting assignments in a cost-effective manner. Also, auditors compete on the basis of reputation, which is only acquired by a history of credible auditing. Compromising audit quality, especially when it allows a company to present rosy results, jeopardizes the auditor s reputation. A second important reason for auditors to value credible and independent audit is the threat of litigation. The liability at both the federal and state levels and requirements of various government agencies with threat of expensive litigation for failure (and potential criminal sanctions) provide strong incentives for auditors to remain 2 See SEC issue Final Rule: Strengthening the Commission's Requirements Regarding Auditor Independence 2/6/ How clearly and what the SEC defines as audit fees, audit-related fees, tax fees, and all other fees has become a contentious issue as far as how informative the required disclosures will be to the market. See The Wall Street Journal, 4

5 independent and vigilant. 4 On the other hand, it has been argued that the independence of an audit is compromised when the auditor believes large consulting fees may be at stake. 5 This study attempts to empirically examine the relationship between an audit failure alleged in a securities class action lawsuit and the various elements of auditor compensation. We find that for our sample of 100 firms in which audit failure has been alleged in a securities class action lawsuit, the audit fees, as well as nonaudit fees are no higher than for a matched sample. 6 However for our sample of firms with the largest market reaction to the audit failure during the period over which the alleged fraud occurred (the class period ), the nonaudit fees was higher than comparable firms. This difference persists even after controlling for known determinants of auditor compensation. The remainder of this paper is organized as follows. Section 2 discusses the extant theoretical and empirical evidence on auditor independence. Section 3 describes our sample selection process and reports descriptive statistics for the sample firms. Section 4 presents evidence on the differences in auditor compensation between our sample firms and the matched sample. Section 5 concludes. 2. Auditor Independence 2.1 Importance of Auditor Independence Proposal May Blur `Audit Fees' --- Plan by the SEC to Tweak Disclosure Rules Could Aid Big U.S. Accounting Firms, January 23, See e.g. DeAngelo (1981), Watts and Zimmerman (1983), Goldman & Barley (1974), Craswell et al. (2002) 5 See Simunic 1984, Parkash and Venable 1993, Firth Our matching algorithm required that each matched firm have total assets within 90% and 110% of the total assets of the litigated firm and have the same SIC code (either 4,3,2, or 1-digit code). Then, we chose the firm with the closest fiscal year end to the litigated sample. Alternatively, we could have chosen our sample of matched firms based on the same SIC code and then matched on total revenue. Previous research shows that size is by far the biggest determinant of audit fees. Therefore we chose to first match on total assets (within a confidence interval) in an effort to hold constant the effect of firm size. 5

6 For capital markets to function well, it is essential that investors are able to get a scorecard on how companies are performing. The scorecard is presented in the form of financial statements that are standardized across companies and follow certain guidelines. The only way in which financial statements are useful to investors is when they are credible. Auditors play an important role in ensuring that accounting statements follow the generally accepted guidelines and are accurate. Expectation is that auditors are independent and will detect and reveal any material omissions or misstatements in the financial statements. That auditors are perceived to be independence is also important to auditors. Auditors can attain credibility with market participants by bonding enough wealth to make dishonest behavior improbable. 7 One important bonding source is the auditor s reputation, which makes the audit credible for investors. Auditors compete on the basis of reputation, which is only acquired by a history of credible auditing. Compromising audit quality, especially when it allows a company to present rosy results, jeopardizes the auditor s reputation. Over time investors will respond to systematic bad auditing by discounting the value of the audit and companies with little to hide will be forced to change auditors. A second important source for auditor bonding is the threat of litigation. The liability at both the federal and state levels and requirements of various government agencies with threat of expensive litigation for failure (and potential criminal sanctions) also provide incentives for auditors to remain independent and vigilant. Given the spate of high-profile audit failures recently, the market, the press, the regulators, and the Congress have all questioned whether the mix of policy and legal mechanisms have been effective in ensuring auditor independence. 2.2 Compensation and Auditor Independence 7 Leland and Pyle (1977), Campbell & Kracaw (1980) 6

7 The debate in the academic literature on the effects of the various components of auditor s compensation on auditor independence has produced mixed results. Some arguments in the literature support the notion of a positive association between auditor independence and compensation. On the other hand, economic dependence of the audit firm on a client may also increase the likelihood that the auditor will acquiesce to management s requests leading to lower quality financial statements. For example, Simunic (1984) models the joint demand for both audit and nonaudit services. He demonstrates that when the auditor provides both services a cost savings (due to knowledge spillovers ) from the joint supply of these services occurs. As a result, when the same auditor provides both services, the cost savings may benefit the accounting firm. The auditor, now earning rents, faces a higher marginal expected loss from being dismissed by top management producing a greater incentive for the auditor to conceal bad news or comply with management. Another view in this literature holds that providing consulting services does not hinder auditor independence and in some cases may enhance auditors incentives to stay independent. DeAngelo (1981) concedes that increased revenues generated by auditors from consulting fees may create an incentive for auditors to compromise their independence and report favorably in order to retain clients. However, when auditors have more than one client there is less financial dependence on a single client. Reputational penalties constrain the behavior of audit firms because the gains from acquiescing to any one client s demands are outweighed by the reputational losses that would be imposed by other clients who need and value the audit firms with a reputation for independence. Similarly, Goldman and Barlev (1974) argue that consulting services combined with auditing services may create a situation in which the client s dependence on the auditor increases because these services enhance the auditor s uniqueness and thus the value to the client. 7

8 In Table 1 we summarize the empirical research that examines the association between auditor independence (and/or financial reporting quality, audit quality) and economic dependence. Since neither auditor independence nor economic dependence is observable, researchers have used several different variables to proxy for these variables. Table 1 displays over seven different measures used by researchers to gauge the extent of the auditor s dependence on the client. Depending on the proxy, the results of these studies have led to conflicting conclusions. Craswell et al. (2002) uses the auditor s propensity to qualify the audit as a measure of auditor independence. For a sample of Australian firms they find that fee dependence does not affect the auditors propensity to qualify their audit opinion (both at the national market level and the local market level). Francis and Reynolds (2001) test the hypothesis that fee dependence will cause auditors to be more lenient and give clients greater discretion in accounting for accruals (both discretionary and total accruals). Surprisingly, they find a negative association between fee dependence and the level of accruals and suggest that reputation protection and litigation avoidance are sufficient incentives for auditors to maintain objectivity. They also find that larger clients (for whom auditors presumably have greater fee dependence) are more likely to receive a going concern audit report. Similarly, DeFond et al. (2002) analyzed financially distressed firms and found no association between audit fees and the propensity to issue a going concern audit opinion. Some studies have shown a positive association between measures of economic dependence and auditor independence (audit quality). Firth (1997) and Parkash and Venable (1993) show that high agency cost firms (determined by lower levels of management ownership, lower outside investment concentration and higher debt) recognize the potential for perceptions of independence impairment and voluntarily limit ex ante purchases of consulting services from the auditors. Their results suggest that auditees recognize the potential for perceptions of independence impairment and higher agency cost firms voluntarily limit the purchase of nonaudit service. Frankel et al. (2001) 8

9 use the level of discretionary accruals and the ability of the firm to just meet or beat earnings targets as a proxy for financial reporting quality. They show a positive association between two measures of nonaudit services and earnings quality. 2.3 Auditor Independence in the Enron Case The Enron audit failure has highlighted the importance of auditor compensation and independence in a very dramatic manner and captured the attention of the market, the press, policymakers and the US Congress alike. On October 22, 2001 a complaint was filed alleging violations of Sections 10(b) and 20(a) of the Securities Exchange Act of 1934 against Enron and three Enron directors. The complaint was filed in response to the disclosure on October 17, 2001 that Enron will take $1.2 billion write-down of its net worth to account for transactions involving related party transactions controlled by CFO, Andrew S. Fastow. On the same day, Enron froze the assets in its 401(k) retirement plan to allow for administrative changes and by the time employees could sell shares, the stock had collapsed. On October 22, 2001, the Securities and Exchange Commission (SEC) requested that Enron voluntarily provide information regarding billions of dollars in certain related party transactions connected to its former CFO, Andrew S. Fastow. On November 11, 2001 the SEC expanded investigation to include Enron s accounting firm, Arthur Andersen. Then, on November 13, 2001 an amended class action complaint for violations of the federal securities laws was filed in which Arthur Andersen along with Enron and nine company officials were named as a defendants. On January 10, 2002, Arthur Andersen disclosed that it had destroyed documents related to work done for Enron. On March 14, 2002 the justice department announced that Andersen had been charged with obstruction of justice in connection with Enron and on June 15, 2002 Andersen was found guilty of the charges. Generally Accepted Auditing Standards (GAAS) as approved and adopted by American Institute of Certified Public Accountants (AICPA), relate to the conduct of individual audit 9

10 engagements and statements on Auditing Standards (AU) are recognized by the AICPA as the interpretation of GAAS. Pursuant to AU , when the auditor examines certain related party transactions the auditor must obtain an understanding of the business purpose of the transaction. These guidelines clearly explain that the audit is not complete until the auditor fully understands the purpose behind the related party transaction. According to AU , Enron s complex partnerships should have been consolidated into Enron s financial statements between the years , not incorporated and restated in In Arthur Andersen s Houston office was engaged to analyze and opine on Enron s financial statements, to perform review services on Enron s interim 2001 results, and to provide consulting, tax and due diligence services throughout 1997 through The complaint states: As a result of the far reaching scope of services provided by Arthur Andersen, they were intimately familiar with Enron s business, including its business relationships. Arthur Andersen received huge fees for its services to Enron. These fees were particularly important to the partners in Andersen s Houston office as their incomes were dependent on continued business from Enron. For 2000 alone, for example, Andersen received $25 million in fees related to the audit of Enron s financial statements and another $27 million nonaudit related work. The complaint continues by quoting a passage from Platt s Oilgram News: Skeptics say those huge fees, and the domination of AA s audit team by Enron s bonus-driven pros, has given Enron great leeway in setting its curve, and thus booking profits. It is widely believed that Arthur Andersen s independence was impaired in the Enron case. It has also been alleged that Arthur Andersen purposely or naively overlooked the accounting for special purpose entities. The fact that Andersen destroyed documents and eventually was found guilty of obstruction of justice is especially troubling. 10

11 3. Sample Selection and Description 3.1 Audit Failure as a Proxy for Auditor Independence The consumers of financial statements expect that an audit would detect and reveal any misstatements of financial information. In some instances, when the audit fails to detect any error in the financial statements and the consumers of that information incur losses, they pursue legal action against auditors and management of the firm. We identify instances in which the consumers of audits (investors) initiate legal action as a consequence of alleged accounting improprieties. These instances are our proxy of audit failure. 8 A caveat about our proxy is in order. Litigation against firms for alleged accounting violation helps us to identify firms for which lack of auditor independence has been alleged, not proven. Some of these securities class action lawsuits are dismissed, and most are settled Sample On November 15, 2000 the SEC adopted revised auditor independence rules requiring firms to separately disclose the amount of audit fees, nonaudit fees and systems design and implementation fees billed by the auditor for the most recent year. 10 Such disclosure permits us to collect auditor compensation data for our sample from firms annual proxy statements. Our sample begins with the universe of firms that where the subject of a class action lawsuit filed during 2001 or We identified such firms through the Stanford Securities Class Action Clearinghouse Database. We found 691 such complaints. The sample is further restricted to those 8 We define audit failure in the same manner as Palmrose (1988) where audit failure includes firms in which there were material misstatements whether or not auditors have obtained sufficient audit information. 9 While there are very few trials, significant settlements are often paid by firms and their auditors in such cases. See Bajaj, Mazumdar and Sarin (2003). 10 Even though the SEC required firms to disclose the amount of Financial Information Systems Design and Implementation Fees in the proxy statement, we exclude such fees from our measure of non-audit fees because Ernst and Young sold their information technology consulting business in May 2000 and that creates problems in comparability across auditors. 11

12 firms with data on the amount of audit and nonaudit fees available in the proxy statements. We further restrict the sample by excluding 14 financial institutions (SIC ) because the relationship between accounting numbers (specifically total assets) and the level of fees paid to a firms auditor may be very different for financial firms as opposed to those included in the sample. Additionally, we limit our sample to those firms that had litigation in connection with an accounting violation. Next, we require the firms to have data on audit fees and nonaudit fees in the proxy statements that overlap the alleged class period or the financial statements in question. 11 We also require our sample firms to have a matched firm, based on industry and total assets (discussed later). This selection process resulted in 100 firms. We obtain data on firm characteristics from Compustat and data on firms acquisition activity from SDC. Panel A of Table 2 provides details on the nature of the main alleged accounting impropriety in the complaint. An examination of the complaints indicate that the nature of the allegations vary widely, however the most common accounting impropriety, in case of 44 firms, relate to certain aspects of revenue recognition. Other common allegations include dealing with asset impairment, expense recognition and debt and/or off balance sheet disclosures. Panel B provides details on how many firms subsequently restated their earnings for at least 1 quarter overlapping our sample period over which we collected data on audit and nonaudit fees. Of the 100 firms, 54 firms have restated earnings or plan to restate earnings and 16 firms have filed for bankruptcy (Panel C). In order to determine whether audit and/or nonaudit fees are higher than expected for firms for which there is an alleged accounting allegation, we construct a matched sample based on SIC code and size of the firms as measured by its total assets. For each firm in the litigation sample, a matched firm was obtained by first identifying all firms in Compustat with an identical SIC code. Out of these firms with a matched SIC code, we select the firm with total assets (measured at the 11 Of our sample of litigation firms, 7 did not have fee data overlapping the class period. 12

13 fiscal year end overlapping the litigated firm s class period) closest in size to the total assets of our litigated firm to be included in our matched sample. However, we require the total assets of the matched firm to be within 90% and 110% of the total assets of the litigated firm and the firm must have the required audit fee data available in its proxy. If no firm is within this size interval, we repeat the process but increase the sample of potential matches by identify all firms with a similar three-digit SIC. If this does not produce a match we identify firms with a similar two digit SIC code, then, if necessary, by one digit SIC code. Out of our sample of 103 litigated firms that otherwise passed our selection criteria, we could identify 100 firms with a matched firm (our selection process did not identify a match for three of our litigated firms). Our matched sample consists of 33, 20, 31 and 16 firms that were match by 4, 3, 2, and 1 digit SIC, respectively. In Table 3 we present descriptive statistics for our sample on certain variables of interest and Analysis of Variance (Wilcoxon) test of differences in the means (medians) between the litigated firms and the matched firms. The litigated firms have an average total assets of billion and median total assets of 878 million. The matched sample has a similar size as measured by the mean (10.69 billion) and median (876 million) of total assets. Tests for the differences in means and medians of total assets between the two samples indicate no significant difference. Similarly, all of the variables presented in Table 3 have means/medians that are not significantly different across the two samples except for two. The mean of foreign tax to total assets and the mean of the indicator variable set equal to one if the firm switched auditor from the previous year are significantly different at a 10% level. (Litigation set was more likely to have changed auditor in the preceding year [in seven percent of the cases], versus the matched sample of firms [about 2% of the cases].) Panel B of Table 3 presents the frequency by industry. The most represented industry is computers, 23%, and next is durable manufacturers at 22% of the sample. Overall the results in Table 3 indicate that the litigated and matched sample are similar. Given this similarity, we would also 13

14 expect the amount of audit and nonaudit fees billed by the firm s auditor to be similar between the two samples. 4. Audit Failure and Auditor Compensation 4.1 Difference in Auditor Compensation for Litigated Firms Table 4 reports the results of t-tests for differences in means (medians) in auditor compensation between firms that had an alleged accounting failure and a matched sample of similar sized firms in the same industry. To study whether results could be different for larger versus smaller firms, we sort our samples into three subsamples based on the value of total assets at yearend, corresponding to the sample year-end for which the audit fee data was collected. Subsample 1 contains the smallest firms and Subsample 3 contains the largest firms. The asset size of the smallest group is 137 million and the largest group is 30,599 million. Consistent with the findings that larger firms require more audit services, we observe that both the mean and median audit fees monotonically increases with size. Indeed, the largest asset size category for the litigated firms has a mean of 3.34 million in total audit fees which is about fifteen times larger than the average audit fees for the smallest firms. A similar pattern holds when we compare nonaudit and total fees. Furthermore, the influence of size on audit and nonaudit fees is also observed in the matched sample. Overall, for the sample of litigated firms, the mean (median) total auditor compensation is 5.3 (1.3) million. Panel A shows that the average audit fees are a little higher for the litigated firms than in the matched sample for all but the smallest firms, however the difference is not statistically significant. In fact, mean (median) total audit compensation between the samples is quite similar, 1.36 (.40) million for the litigation sample and 1.31 (.51) million for the matched sample. The remaining four panels reveal that the mean and median for nonaudit fees, total fees, nonaudit fees as a fraction of total fees and as a fraction of audit fees, respectively, is somewhat higher for the firms involved in 14

15 litigation, but the difference is not statistically significant. The results do not provide support for the hypothesis that firms involved in an accounting impropriety had relatively higher audit compensation. Table 5 reports means and medians of audit and nonaudit fees by industries for which there are at least 10 firms represented by the sample (Computers, Durable Manufacturers, Services, Utilities). Panels B and C report total fees by industry and show that only the utility industry has higher nonaudit fees and total fees that are weakly significant at a 10% level. However, for the other three industries the fee (audit, nonaudit and total) differences are statistically insignificant. Since only the utilities industry shows a weakly significant difference across the two samples, the results provide little support for the hypothesis that the fees paid to a firm s auditor impair auditor independence or effectiveness. In Table 6 we examine two sub-samples of the litigation sample. In particular, we analyze the sub-sample of firms that subsequently filed for bankruptcy and firms that subsequently restated earnings. Our results reveal no significant difference between the litigation firms and the matched sample firms. Panel A reveals that the 16 firms in the litigation sample that subsequently went bankrupt did not pay their auditors more relative to the matched sample. Interestingly, the mean nonaudit fees for the litigation sample (2.2 million) is almost double that of the matched sample (1.2 million), however the difference is not significant. Firms that report financial statements that are later determined to be inaccurate or fraudulent oftentimes are required to restate earnings. A report prepared by the U.S. General Accounting Office (GAO) found that 10 percent of publicly traded companies restated financial statements because of accounting irregularities from January 1997 to June "In a number of the restating companies we identified, corporate management, boards of directors, and auditors failed in 15

16 their roles, as have securities analysts and credit rating agencies that did not identify problems before investors and creditors lost billions of dollars," the study said. However, Panel B shows that the mean/median difference between the fees from the litigated sample and the matched sample are not significantly different for any fee component. The mean (median) audit compensation between the samples is quite similar, 1.4 (.56) million for the litigation sample and 1.3 (.65) million for the matched sample. The mean nonaudit compensation between the samples is larger for the litigation sample but the median is larger for the matched sample. However, not all restatements necessarily represent serious misstatements. GAAP do not always provide a uniquely correct answer on accounting treatment, nor would a rigid system of rules be always consistent with the objectives underlying GAAP. As the importance of intangibles and complexity of businesses grows, the scope for ambiguities in GAAP grows as well. Moreover, after high-profile audit failures in the Enron and other cases, auditors and firms may well be restating results in an effort to be more conservative in the current environment. Therefore, it is not necessarily the case that all restatements represent significant audit failures. In an attempt to quantify the severity of the audit failure we calculated change in value of the firm s common equity from the day with the highest stock price in the class period to the day after the class period. Firms with the greatest decline in stock price are assumed to represent the most severe audit failures in market s judgment. 4.2 Severity of the Alleged Audit Failure Our sample of litigation firms are those in which audit failure has been alleged, not proven. In fact, many shareholder class actions are dismissed by courts before any violation is proven. 12 Financial Statement Restatements: Trends, Market Impacts, Regulatory Responses, and Remaining Challenges, 16

17 Almost all of the rest are settled out of court. Clearly, class action lawsuits are filed in many cases even when no audit failure has occurred. To improve the signal to noise ratio for our proxy, we sort our sample of litigation firms based on the magnitude of the stock price decline between the highest stock price during the class period and the end of the class period (which is usually when corrective disclosure is made). The interpretation of our proxy for audit failure could be confounded by the issue that class action laws may provide incentives to sue firms with negative abnormal returns preceding the class period 13. It may be the case that a large decline in stock price leads to the initiation of a lawsuit. However, given this possible limitation, we believe our proxy does do a good job capturing the most severe audit failures in market s judgment. Table 7 reports the results for a third of our firms with the largest decline in market value, the middle third and the one s with the least decline in market value of their stock. The group of firms with the highest market reaction lost on average 81% of their value while the firms with the lowest market reaction lost 25%. Our findings in Panel A indicate that, for firms with the most severe audit failures, audit fees are no different between the litigated firms and their matched sample. However, the litigated firms have significantly higher nonaudit fees. The differences in nonaudit fees measured as a fraction of total fees and as normalized by audit fees, are higher for the litigation subsample, but the p-values for the difference would make it significant at the 10% level in one-tailed test only. The mean (median) difference between the fees from the litigated sample and the matched sample is 10.7% (10.2%) and 173.4% (73.1%) for the fraction of nonaudit fees to total and fraction of nonaudit fees to audit fees, respectively. The findings in Panel B and Panel C, for the middle third and the least severe audit failures the test statistics show there are no statistically significant differences for all fee components except in one case. The fraction of October See Kellogg (1984) 17

18 nonaudit fees to total fees for the least severe audit failure are statistically significant at a 10% level but in the direction opposite to that suggested by the fee dependence argument. 4.3 Multivariate Analysis In the appendix to this paper, we develop a model to explain the cross-sectional differences in for the natural log of total fees, audit fees, nonaudit fees and two ratio measures of consulting fees: the ratio of nonaudit fees to total fees and the ratio of nonaudit fees to audit fees. Several determinants of fees paid to a firm s auditor have been well documented in the academic literature. Our model has similar explanatory power to other fee structure models used the prior literature 14. Our results are similar whether we use the litigation, matched or combined sample. 15 The appendix provides details on the empirical estimation and results of the model. Table 8 presents the results from the cross-sectional regressions relating audit compensation to an indicator variable equal to one if the firm is in the litigation sample and various control variables. The litigation indicator dummy is the variable of interest used to test the conjecture that firms involved in an audit failure paid relatively higher compensation to their auditors. We also include an indicator variable equal to one if the firm restated earnings. The log of total assets, the ratio of foreign income taxes to total sales and return on assets all significantly explain the level of audit fees. However, our results indicate that, after controlling for other factors that may explain fees, the difference in fees between the litigation sample and the matched sample are not statistically different. Our variable of interest, the indicator variable for the litigation sample, is not significant at explaining the level of audit, nonaudit, total fees or the ratio of consulting fees: the ratio of nonaudit fees to total fees and the ratio of nonaudit fees to audit fees. 14 See for example: Craswell et al. (1995), Seetharaman et al. (2002), Firth (1997), Parkash and Venable (1993) 15 See Appendix, Tables A-1, A-2 & A-3 18

19 The results from Table 8 are not consistent with the hypothesis that auditors are overpaid (audit and/or nonaudit) fees in order to overlook various GAAP rules and acquiesce to management s requests leading to lower quality financial statements. However, statistical inference on the litigation indicator variable could be affected by multicollinearity. Multicollinearity could exist if there is a large correlation between our indicator variable of interest and any of the control variables. The largest correlations among the control variables are between the litigation indicator and foreign tax to total sales ( -.12), the change indicator (+.12) and the acquisition indicator (+.11), with all other correlations less than Table 9 presents the results of the same cross-sectional regressions presented in Table 8, but we limit the sample to the third of our firms with the largest decline in market value. Our results indicate that for these firms with the most severe audit failures, the three measures of nonaudit fees (Model 2, Model 4 and Model 5) are significantly different, at a 1% level, between the litigated firms and their match. In contrast, there is no significant difference between the litigated and match sample in explaining the cross-sectional variation in the log of audit fees. The results (not presented) for the middle third and the least severe audit failures show no significant difference between the two samples 16. Taken together, the results from Table 9 are consistent with the hypothesis the auditor s independence may be compromised in their bid to attract consulting business from their auditing clients. However, given the small size of our subsample (66 firms) we can not make a conclusive argument. 5. Summary and Implications Record number of audit failures during the recent past has prompted much debate in the popular press, policy circles and in the US Congress about whether high auditor compensation, especially for nonaudit work, may have led to lax auditing standards. We shed light on this question by 19

20 comparing auditor compensation for a set of firms in which accounting improprieties were alleged in a shareholder class action lawsuit with a set of matched firms in the same industry and of similar size. Our evidence suggests that auditors were not compensated differently for either their audit or consulting services over the period in which their client was allegedly involved in an accounting fraud. However, for the set of firms with the largest market reaction to the alleged fraud, the nonaudit component of the total fees was significantly higher then comparable firms even after controlling for other known determinants of auditor compensation. While our analysis documents that nonaudit fees are indeed higher than normal in cases for which there was a severe audit failure, this result should be interpreted with caution. There may be perfectly valid business reasons for companies to use the auditors for consulting activities more than similar sized firms in the same industry. Therefore, higher compensation for consulting activities cannot be interpreted as evidence of lack of auditor independence. Any allegation of auditor integrity being compromised can only be made based on analysis of facts and circumstances. 16 Results available upon request 20